Академический Документы
Профессиональный Документы
Культура Документы
=
N r
N r
r
r
r
r
LS
X
Y
X
Y
X
Y
G , ,
2
2
1
1
Where
r
N
LS
C G e the estimated channel frequency response
on the sub carriers which contains reference symbols. This
response can be interpolated over full frequency range in
order to obtain the channel frequency response for the
subcarriers carrying data symbols. The interpolation can be
performed in time domain or frequency domain.
The signal received in time domain can be expressed as
follows:
+ = h F A F Y
L r
H
r
The channel can be estimated using Least Squares, in time
domain in the following
r
H H
Y S S S h
1
) (
=
Solving the above two equations, we get the expression for LS
estimates.
r
H H
R r
H
L L r
H
L
H
L
Y F A Y F F A A F h
1
) (
~
Regularized LS Estimation
International J ournal of Computer Trends and Technology (IJ CTT) volume 5 number 1 Nov 2013
ISSN: 2231-2803 http://www.ijcttjournal.org Page52
a
In this method, small constant term is added to the diagonal
entries for regularizing the Eigen values of the matrix to be
inverted. In this case, the channel impulse response becomes
of the form,
r
H H
R r
H
L L r
H
L
H
L reg
Y F A Y F F A A F aI h
1
) (
+ =
The value of has to be selected such that the inverse matrix
is least perturbed.
Minimum Mean Square Estimation
LS channel estimation method has been described which is
computationally simple but its performance is not good.
Another method to estimate the CIR is minimum mean square
estimator (MMSE) which has better performance than LS but
it is computationally complex. This method intends at the
minimization of the mean square error between the exact and
estimated CIRs. In this section we will discuss linear
minimum mean square estimator (LMMSE). The CIR can be
calculated using LMMSE in the following way
r y y hy
Y R R h
r r r
1
=
Here
r r
y y
R is the auto covariance of vector
Yr
and R
hyr
is the
cross covariance of vectors h and Y
r
. These covariance
matrices for the above equation can be calculated as,
Nr
2 H
r
H
r hh r r
H
r
H
r
I T X R T (X T X = h
o +
Fundamentals of turbo codes:
In order to explain the proposed parallel Turbo decoder
architecture, the fundamentals of Turbo codes are briefly
described in this section.
Turbo encoder structure:
As shown in Fig. 4, the Turbo encoding scheme in the LTE
standard is a parallel concatenated convolution code with two
8-state constituent encoders and one quadratic permutation
polynomial (QPP) inter leaver. The function of the QPP inter
leaver is to take a block of N-bit data and produce a
permutation of the input data block. From the coding theory
perspective, the performance of a Turbo code depends
critically on the inter leaver structure. The basic LTE Turbo
coding rate is 1/3.It encodes an N-bit information data block
into a codeword with 3N+12 data bits, where 12 tail bits are
used for trellis termination. The initial value of the shift
registers of the 8-state constituent encoders shall be all zeros
when starting to encode the input information bits.
Turbo decoder structure:
The basic structure of a Turbo decoder is functionally
illustrated in Fig. 5. A Turbo decoder consists of
two maximum posterior decoders (MAP) separated
by an interleave that permutes the input sequence.
Fig.4: Structure of rate 1/3 Turbo encoder in LTE.
Each Turbo iteration is divided into two half
iterations. During the first half iteration, MAP
decoder 1 is enabled. It receives the soft channel
information (soft value L
s
for the systematic bit and
soft value L
p1
for the parity bit) and the a priori
information L
1
from the other constituent MAP
decoder through de-interleaving (
) to generate
the extrinsic information L
1
at its output.
Fig 5: Basic structure of an iterative Turbo decoder. (a) Iterative
decoding based on MAP decoders. (b) Forward/backward
recursions on the trellis diagram.
Likewise, during the second half iteration, MAP
decoder 2 is enabled, and it receives the soft channel
information (soft value L
s
for a permuted version of
the systematic bit and soft value L
p 2
for the parity
bit) and the a priori information L
2 from
MAP
decoder 1 through interleaving to generate the
extrinsic information L
e
at its out put . This iterative
process repeats until the decoding has converged
International J ournal of Computer Trends and Technology (IJ CTT) volume 5 number 1 Nov 2013
ISSN: 2231-2803 http://www.ijcttjournal.org Page53
o r the maximum n u mb e r of iterations has
been reached. The MAP algorithm at each
constituent MAP decoder computes the log-
likelihood ratios (LLRs) of the a post er i or i
probabilities (APPs) for information bit uk
.
MAP algorithms
Corresponding to different contents of encoder memories the
encoder can be in one of 2
M
States for the binary input of
alphabet {1,-1}. The trellis structure is exploited to compute
the a posteriori L-values
=
=
=
=
=
=
1 ), , (
'
1 ), , (
'
,
,
) , (
) , (
log
) 1 (
) 1 (
log ) (
xk s s
k
xk s s
k
k
k
k
s S
s S
y x p
y x p
x L
Where s 'and s represent the states of the encoder element at
time k-1 and k, and y are the received sequence of length N,
respectively. Dividing the received sequence into three
separate terms and applying the Bayes chain rule, we get
1 , 1
'
1
1 , 1
'
( ). , (
) ( ) , (
+
+ +
= =
= =
k k k
k n k k
y s p s sk s s P
s s y p s s
Here using forward or backward algorithm the above equation
can be evaluated as 3 terms
) , ( ) , (
) ( ) (
) , ( ) (
1
'
, 1
1 , 1
k k k
k n k k k
k k k
y s s p s s
s s y s p s
y s s p s
= =
= = =
= =
+
+
|
o
Using the forward and backward recursions
()
() are
computed as
()
()
()
()
()
()
()
Here the important is the calculation of the
) is
different for both the encoder and decoder, since the existence
of treelis codes.
For the equalizer
) (
))
For decoder
) (
))
Over AWGN channel the turbo decoding without the use of
Inter State Information
(
).
Using the extrinsic Jacobian logarithmic function The log-
Map algorithm introduced to evaluates the
(s) and
(s)
and
(s, s').
() Max [
)]
() Max [
)]
Where
) ( (|
|))
(s),
(s), and
(s)
and
(s) and
(s, s'), respectively.
MAP log likelihood ration of the decoder is determined as
follows:
L (
) =
(
()
)]
()
)]
In the trellis structure max operation is performed in first and
second part only .the main advantage of the Max-Log-Map is
ignores the interference caused by the SNR.
V. SUMULIATION AND PERFORMACE ANALYSIS
In this paper we, Matlab is used to realize the simulation
experiment.
Fig3: Comparison of BER for various estimation techniques.
International J ournal of Computer Trends and Technology (IJ CTT) volume 5 number 1 Nov 2013
ISSN: 2231-2803 http://www.ijcttjournal.org Page54
B. Simulation results:
In Fig3, We can see that the performance of the LS estimation
is worse in the fading channel. The reason for the low
estimation accuracy is that we have little reference symbols to
interpolate the channel response. After reconstructing the
original transmission OFDM symbols. It means that use more
system signals as reference symbol; the performance of the
iterative LS estimation could near the MMSE estimation. As a
comparison, the BER curve of perfect channel estimation is
depicted in the Fig. The simulation result shows that through
the proposed iterative channel estimation method, dramatic
performance improvements can be achieved with the LS
estimation and that the advantage of low implementation
complexity makes it possible widely used in the practical
OFDM system.
REFERENCES
[1] Luis Angel Maestro Ruiz de Temino and Carles Navarro I Manchon,
Iterative Channel Estimation with Robust Wiener Filtering in LTE
Downlink,IEEE Trans.Commn.,Aug, 2008
[2] Andrea Ancora , Calogero , Bona Dirk T.M. Slock , Down-Sampled
Impulse Respponse Least-Squares Channel Estimation For LTE OFDMA,
IEEE Trans.,ICASSP 07.
[3] Farzad Foroughi , Johan Lofgren, Ove Edfors,Channel estimation for a
Mobile Terminal in a Multi-Standard Enviroment (LTE and DVBH). IEEE
Trans. Aug 2009.
[4] Members of 3GPP, Technical Sepcification Group Radio Access
Network;Physical Channels and Modulation (3GPP TS 36.211 Version
8.8.0),3GPP,Tech.Spe.,Sep,2009..
[5] L. Bahl, J. Cocke, F. Jelinek, and J. Raviv, Optimal decoding of linear
codes for minimizing symbol error rate, IEEE Trans. Inf. Theory, vol, IT-20,
no. 2, pp.284-287, Mar.1974
[6] G. Bauch, H. Khoram, J. Haganauer, Iterative equalization and decoding
in mobile communications systems, in Proc. 2nd EPMCC, 3.ITGFachtagung
Mobile Kommunikation, Sep./Oct. 1997, pp. 307
312.VDE/ITG..
[7] P. Robertson, P. Hoeher, and E. Villebrun, Optimal and sub-optimal
maximum a posteriori algorithms suitable for turbo decoding, Eur. Trans.
Telecommun., vol. 8, no. 2, pp. 119125, 1997.
[8] J. A. Erfanian, S. Pasupathy, and G. Gulak, Reduced complexity
symbol detectors with parallel structures for ISI channels, IEEE
Trans.Commun. , vol. 42, no. 234, pp. 16611671, Feb.Apr. 1994.
Mathe Sowmya currently pursuing her M.Tech degree in Sir C.R.Reddy
college of Engineering, Affiliated to Andhra University. She was graduated
from JNTU Kakinada in the year of 2011.
CH. Ravi Kumar, currently working as an Assistant Professor in Sir
C.R.Reddy Engineering College, Eluru. He has done much amount of work
towards the new concepts in the era of turbo codes and pursuing Ph.D. degree
from JNTU Kakinada. He was post graduated in M.Tech from JNTUK
Kakinada and graduated in B.E in SRKR Engineering College, with the both
specializations are in Electronic and Communication Engineering.
[1] .